The progression of chronic renal failure is relentless in the majority of patients with chronic nephropathies. The extent of tubulointerstitial fibrosis correlates best with the deterioration in glomerular filtration rate in all kidney diseases including hypertensive nephropathy (HN).1, 2, 3, 4 This is due to the underlying scarring process characterized by the depletion of kidney cells and their replacement by extracellular matrix (ECM), in particular by type-I fibrillar collagen.5 Normal kidney structures express very little collagen I in the interstitium and glomeruli, whereas renal and extrarenal arteries constitutively express adventitial collagen I. Although most investigations on HN have focused on the glomerulus and renal microvasculature, tubulointerstitial injury is also a major feature of HN, and remains an important predictor of renal dysfunction. This is striking because HN is generally attributed to increased mechanical stress of the vasculature, which may explain vascular adaptative fibrosis similar to that observed in extrarenal arteries, but may not explain the heterogeneous spreading of tubulointerstitial fibrosis. In addition, several recent experimental and clinical studies provided evidence that tubular epithelial cells can undergo phenotypic change, acquiring myofibroblast-like features such as smooth muscle actin expression, loss of cytokeratin, and de novo fibrillar collagen I and III expression,6 thereby undergoing epithelial-mesenchymatous-transition (EMT).7, 8 Thus, several potential cell types may promote interstitial fibrosis, and it is therefore crucial to develop tools to understand where collagen fibers originate and accumulate in the kidney during pathologies.
Two bottlenecks have impaired efforts of investigators to quantify and locate kidney fibrosis. First, interstitial collagen fibrils form elaborate 3-D arrays that extend over micrometers to millimeters. These 3-D networks are therefore hardly quantified by contrast with ECM accumulation in well-defined structures such as glomeruli and vessels sections. Second, conventional histological staining techniques (hematoxylin and eosin, periodic acid Schiff or Masson’s trichrome stains) are not specific to fibrillar collagen. Specificity is improved using picrosirius red with polarized light9 or immunohistochemical labeling; however, these techniques lack 3-D resolution. In addition, since assessment of the fibrillar collagen is heavily dependent on the thickness of histological sections and the yield of staining, conventional stains lack reproducibility for quantitative purposes. For instance, evaluation of interstitial fibrosis during chronic allograft nephropathy is the most inconsistent index in the Banff score, which is the standardized classification of allograft pathology.10
In this context, we propose to take advantage of intrinsic second harmonic generation (SHG) from fibrillar collagens. SHG is a multiphoton process where an intense impulsive laser source, at frequency , induces a second-order nonlinear polarization in the sample under study, which coherently radiates at the harmonic frequency . SHG signals are therefore detected at exactly half the excitation wavelength, whereas two-photon excited fluorescence (2PEF) signals are red-shifted compared to the SHG frequency. SHG from collagen fibers was first reported in the 1980’s in a rat-tail tendon11 and then utilized in multiphoton microscopy in combination with 2PEF 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25 and third-harmonic generation26 to visualize the 3-D morphology of thick unstained tissues. We recently showed that SHG microscopy presents unique advantages compared to conventional optical techniques for quantitative 3-D imaging of fibrillar collagen.22 First, SHG microscopy is a specific structural probe for fibrillar collagen, and no SHG signal is obtained from type-IV nonfibrillar collagen in tissues.22, 27 The high specificity of SHG signals from fibrillar collagen results in a small background noise in SHG images, enabling sensitive and quantitative measurements of the extent of the fibrillar network. Reproducibility of these measurements is furthermore ensured by the absence of any staining or labeling. Second, SHG microscopy is a versatile technique: it does not require any preparation of the tissue and is applicable to fresh, frozen, or fixed tissues (including alcoholic fixation or paraffin embedding).21 Third, SHG microscopy presents intrinsic 3-D resolution due to its multiphoton character. It is thus insensitive to mild thickness variations in fixed or frozen sections. Most importantly, SHG microscopy is robust upon light scattering within tissues, and enables improved depth penetration within tissues compared to confocal microscopy.
In the present study, we demonstrate that SHG microscopy is a valuable tool for the investigation of heterogeneous accumulation of fibrillar collagen during experimental hypertensive renal fibrosis and human chronic allograft nephropathy, and permits the precise measurement of hypertensive vascular remodeling at the micrometer scale. Furthermore, SHG microscopy allows us to improve the characterization of the hypertensive renal fibrosis induced by Angiotensin II (AngII) in mice and to unravel the spatiotemporal progression of fibrosis in the different renal compartments.
Materials and Methods
Human kidney samples were collected from transplant recipients with stable graft function after engraftment of kidney from a deceased donor. Transplantations were performed at Hôpital Tenon, Paris, France. Needle-core biopsies were obtained using a biopsy gun (Monopty Bard, Covington, United Kingdom) with a 16-gauge needle.
Models of Angiotensin-II-Induced Hypertension and Crescentic Glomerulonephritis
We studied male mice aged with similar genetic background (C57BL/6J background). Baseline blood pressure measurements were performed on five consecutive days. After these baseline recordings, an osmotic minipump (Alzet Model 2004; Durect, California) infusing AngII (Sigma, Saint Louis, Missouri) at a rate of was implanted subcutaneous for as described by Crowley 28 Blood pressure was regularly monitored to verify continuous hypertension. In a subset of experiments, a second osmotic minipump was implanted on day 27 to ensure AngII-infusion up to day 49. The animals ingested a normal diet containing 5% sodium chloride (SDS-Dietex, France) to accentuate hypertension and kidney injury. We also used the accelerated antiglomerular basement membrane anti(GBM) crescentic glomerulonephritis (GN) model, as described by Lloyd 29 Briefly, mice were given subcutaneous injections of normal sheep IgG [ into each flank of normal sheep IgG (Sigma) diluted to in a solution of 50% Freund’s complete adjuvant and 50% saline]. Six days later (day 1), GN was induced by intravenous injection of sheep anti-GBM serum. Serum injections were repeated daily on days 2 and 3. The protocols followed the French Department of Agriculture and the National Institute of Health guidelines for animal care and protection.
Human kidney biopsies specimens were fixed in formaldehyde. After fixation, tissues were dehydrated with a graded series of ethanol and xylene, embedded in paraffin, and cut into sections for SHG analysis.
Following 28 or of AngII infusion or after anti-GBM serum administration in C57BL6/J mice, murine kidneys were harvested and weighed. One half of a kidney was fixed in formalin, embedded in paraffin, sectioned , and stained with Masson’s trichrome. All of the tissues were examined by a pathologist without knowledge of treatment groups. The pathological abnormalities were graded based on the presence and severity of component abnormalities, including glomerulosclerosis. Grading for glomerulosclerosis was performed using a semiquantitative scale as previously described.30 Percent glomerulosclerosis was defined as the number of glomeruli with evidence of sclerosis divided by the total number of glomeruli in the section. For tubulointerstitial areas, individual scores adapted from Spurney 31 were obtained for the severity of interstitial inflammatory cell infiltrates and severity of tubular atrophy. The second half of the kidney was fixed in phosphate buffered saline (PBS) buffered 4% formaldehyde and used for multiphoton microscopy. -thick coronal slices were cut using a vibrating-blade microtome (VT1000S, Leica, Germany) at 25% of the kidney height.
Multiphoton images were recorded using a custom-built upright laser scanning microscope, as previously described.22 Briefly, SHG and 2PEF signals were simultaneously excited by a femtosecond titanium-sapphire laser adjusted to (Tsunami, Spectra-Physics, Irvine, California), and dispatched to two photon-counting epidetection channels using a dichroic mirror (FF458-Di01, Semrock, Rochester, New York) and appropriate spectral filters (E700SP and GG455 in the 2PEF channel, E700SP and HQ430/20 in the SHG channel, Chroma, Rockingham, Vermont). Alternatively, the SHG signal from thin biopsies was detected in the forward direction using the same filters. We used circular polarization to minimize orientation effects in the image plane, similarly to circularly polarized Picrosirius imaging.9
Visualization of micrometer or submicrometer collagen fibers that were heterogeneously distributed in the kidney tissue at millimeter scale required both a large field of view and a good spatial resolution. We therefore used a 0.95-NA objective lens (Olympus, Japan) with field of view and 0.40- (axial) resolution near the sample surface. Furthermore, we scanned the tissue using a motorized stage with travel range. We always scanned areas with pixel size, at pixel rate and excitation, and we acquired eight images spaced out apart within the depth of the tissue. Voxel size was set to a slightly greater value than the optical resolution near the sample surface to restrain imaging time to typically per sample. Nevertheless, images acquired for 3-D reconstructions were recorded with pixel size and were spaced out apart within the tissue.
Second Harmonic Generation Image Processing and Segmentation
Stacks of SHG and 2PEF images were combined and processed using an algorithm developed to enhance the fiber contrast, as previously described.22 3-D reconstructions were performed using Amira (Mercury Computer Systems, Chelmsford, Massachusetts). Fibrosis scores were obtained from SHG data using a phenomenological approach as previously published:22 they were calculated as the volume density of voxels exhibiting significant SHG signal in the image stack.
To take into account the specific morphology of every murine kidney, that is the variability in kidney shape and in radial location of the arteries, we developed a segmentation algorithm and calculated fibrosis scores on distinct morphological regions of the kidney. We first manually outlined the capsule and the papilla in every SHG/2PEF image, and selected the centers of the arcuate arteries in the cortico-medullary area. Our algorithm then yielded three curves: Ca, along the capsule, Ar, through the arterial centers, and PM, at the edge of medulla and papilla [see Fig. 1a ]. These curves were fitted to ellipses, enabling their dilation or contraction toward the center of the ellipse, as shown in Fig. 1b. Three regions were defined by our algorithm [Fig. 1c]: the outer cortical region [C] between and , the medulla region [M] between and , and the papilla region [P] limited by . The contraction of Ca by ensures that large SHG signals from the capsule do not contribute to the cortical SHG score. The shift of the Ar curve apart from the artery centers aims at excluding the arteries from any SHG scoring, because their quantity and size in every kidney sample show large fluctuations (see results). The replicas of MP minimizes the uncertainty of the borderline position between the medulla and the papilla. Our algorithm finally provided SHG scores for all these regions, which were calculated as the percentage of nonzero SHG voxels averaged over the stack.
Quantitative Analysis of the Arcuate Artery Hypertensive Remodeling
We performed a quantitative analysis of the morphology of arcuate arteries. For every mouse, we selected one to five arteries. Their section was approximately elliptical because their orientation in the tissue was not absolutely perpendicular to the image plane. We considered only the small axis of the ellipse, which was not distorted by the 2-D sectioning, and measured: 1. the lumen radius L, as delineated by 2PEF from the internal elastic lamina; 2. the thickness of the media (region exhibiting 2PEF) M; and 3. the thickness of the adventitia (region exhibiting SHG) A. The measured artery radii ranged from (average: ). The mean cross sectional area (CSA) was calculated for each experimental condition. We normalized the data with respect to the media outer radius for every artery to compare the morphology of arteries having different sizes. We then calculated mean values of the raw data (in ) and of the normalized dimensions (in %) for every mouse. The results were averaged within the three groups: the control mice , the -infused mice , and the -infused mice . The same method was used for the calculation of the media to lumen (M/L) and wall to lumen ratios.
The values for each parameter within a group were expressed as the standard error of the mean (SEM). For comparisons between groups, a one-way nonparametric Kruskal-Wallis test was employed followed by Dunn’s Multiple Comparison test (Prism 5, GraphPad Software Incorporated, San Diego, California). For comparison of different measurements performed on the same group, we used the Kendall tau coefficient.
Multiphoton Images of Human and Murine Renal Sections
We first studied human renal biopsies removed three months after renal transplant to assess interstitial fibrosis. Figure 2a shows a multiphoton image of an unstained formalin-fixed paraffin-embedded section: SHG and 2PEF contrasts were recorded simultaneously and displayed with green and red colors, respectively, in the combined image. 2PEF from endogenous cellular chromophores revealed tubules, glomeruli, and arterioles. SHG revealed collagen fibers in the Bowman's capsule and in the tubular interstitium. We also observed small fluorescent structures surrounded by collagen fibers in the interstitium, consistent with capillary morphology.
To study the progression of interstitial fibrosis and to compare the features of mouse renal fibrosis with human renal fibrosis, we used a well-characterized animal model of hypertensive renal fibrosis.28 C57B16 mice were infused for with AngII, and they developed increased systolic blood pressure, albuminuria (data not shown), and glomerulosclerosis compared with normotensive counterparts, as evidenced by Masson’s trichrome staining [Figs. 2b, 2c, 2d]. Serial sections of kidney cortex from the same control and hypertensive mice were visualized using multiphoton microscopy [Figs. 2e, 2f, 2g]. Multiphoton images of representative sclerosed glomeruli [Figs. 2f and 2g] showed numerous collagen fibers in the Bowman's capsule and in the tubular interstitium, whereas the control unstained section [Fig. 2e] showed no SHG signal, thus no fibrillar collagen accumulation, except for a few thin fibers in the tubular interstitium. Notably, as in human kidneys [Fig. 2a], collagen fibers were never observed within the glomerular tuft despite patent glomerulosclerosis after chronic and severe hypertension [Figs. 2f and 2g].
Three-Dimensional Distribution of Collagen Fibers in Murine Fibrotic Kidney Tissue
To characterize the collagen distribution on a large scale, we imaged renal cortex and medulla in thick coronal slices [Figs. 3a, 3b, 3c ]. Figure 3a shows a combined SHG/2PEF image from a -infused mouse obtained by stitching many images acquired using an objective with a high numerical aperture. We observed collagen fibers in the adventitia of arcuate arteries and arterioles and in the tubular interstitium.
We also took advantage of the 3-D imaging capability of multiphoton microscopy to visualize the 3-D distribution of collagen fibers in the adventitia from arcuate arteries [Fig. 3d] and in the tubular interstitium [Fig. 3e]. For this purpose, we reconstructed 3-D images from stacks of combined SHG/2PEF images. Strong 2PEF signals arose from elastic lamina in the artery media as previously observed19, 20 [Fig. 3d]. As expected, SHG revealed collagen fibers in the vessel adventitia [Fig. 3d] and in the tubular interstitium [Fig. 3e], where they appeared to form a quasicontinuous network.
Quantitative Analysis of Arcuate Artery Hypertensive Remodeling
Characterization of the remodeling of resistive arteries in end organs is of major importance in evaluating tissue perfusion, since the peripheral resistance is mainly determined by the distal part of the arterial vasculature (the resistance vessels), consisting of small arteries (diameter ) and arterioles.32, 33 Arcuate arteries are less than in diameter in the mouse kidney, well below the usual threshold of . We focused on these arteries and measured the overall vascular size, the adventitia thickness, the media thickness, and the lumen radius, as shown in Fig. 4a . Table 1 displays raw and normalized results for control and hypertensive mice. Mean raw data did not present significant changes with hypertension, because of the size heterogeneity of the arteries. However, normalization, taking into account artery radius, revealed a decrease in the lumen radius in hypertensive animals and a concomitant increase of the media thickness. By contrast, we observed no significant change in the adventitia normalized thickness [Fig. 4b].
Dimensions of arcuate arteries in control and hypertensive mice. Mean values and standard errors of the mean of the arcuate artery dimensions for control mice versus 28- and 49-days AngII-infused mice. Raw values are directly obtained from length measurements, as schematized in Fig. 4a. Normalized values are obtained as the ratio of the raw data to the media outer radius (L+M) in every artery, as explained in the methods. ( p*<0.05 versus control value, n=7 ).
|Vascular size||Artery radius ( , )|
|Media radius ( , )|
|Cross sectional area||2079||2440||2041|
|Adventitia thickness||Raw value (A, )|
|Normalized [ , %]|
|Media thickness||Raw value (M, )|
|Normalized [ , %]|
|Lumen radius||Raw value (L, )|
|Normalized [ , %]|
These structural properties of arteries helped determine the physiologically important media to lumen (M/L) and wall to lumen ratios [Figs. 4c and 4d]. Hypertension significantly increased media to lumen and wall to lumen ratios in these renal resistance arteries. This altered morphology was not associated with any significant increase in the media cross sectional area (Table 1), suggesting rearrangement of the media around a smaller diameter, so-called eutrophic remodeling.32
Quantitative Analysis of Segmented Second Harmonic Generation Images Reveals Angiotensin-2-Induced Interstitial Fibrosis in the Cortex
We compared the progression of fibrosis in normal and hypertensive mice infused for 28 or . Interstitial fibers were heterogeneously distributed in a thin network that accumulated mostly in the inner cortex, close to arcuate arteries, and extended radially in the medulla and in the outer cortex [Figs. 5a, 5b, 5c ]. Surprisingly, we observed physical connections (or at least very close proximity below the resolution of the microscope) between perivascular, periglomerular, and peritubular fibers. For comparison, we also visualized kidneys from a mouse model of crescentic glomerulonephritis, and we observed a similar distribution of collagen fibers [see Fig. 5d].
To discriminate between de novo interstitial fibrosis and constitutive adventitial collagen around arcuate arteries, we performed a radial segmentation of SHG images excluding arcuate arteries as well as border regions, and selected outer-cortical, medullar, and papillary regions [see white dotted line in Figs. 3c and 5]. We then obtained fibrosis scores in various anatomical regions from the kidney tissue by measuring the density of voxels exhibiting nonzero SHG signal.
Averaged values of the cortical scores and of the total scores are shown in Fig. 6 for control and hypertensive mice. Both scores increased for infused mice, and with the infusion duration. Cortical scores showed the larger relative increase in hypertensive kidneys. Medullar scores were highly heterogeneous in control kidneys and showed no significant change with hypertension (data not shown). Papillary SHG signals were scarce, with low signal over noise ratios and unreliable SHG scores.
Finally, we confronted cortical SHG scores with usual histopathological methods that evaluate kidney damages in Masson’s trichrome stained sections. We plotted cortical SHG scores as a function of histopathological evaluation of interstitial damage and inflammation [Fig. 6b] and glomerulosclerosis scores [Fig. 6c] for all mice in the three groups (control, and AngII-infused mice, mice). Semiquantitative histopathological evaluation of interstitial damage and SHG cortical score were found to be correlated ( , ) [Fig. 6b]. Glomerulosclerosis correlated better with cortical SHG scores ( , ) than with global SHG scores ( , ).
Although the extent of tubulointerstitial fibrosis correlates best with the deterioration in glomerular filtration rate in kidney diseases including hypertensive nephropathy,1, 3, 4 quantification of 3-D arrays of collagen fibers remains a challenge in research and clinical practice. Hypertension is a prominent cause of end-stage renal disease. Yet it is not fully elucidated whether interstitial fibrosis is an early event or not, and where collagen fibers do assemble in the context of causative vascular stress. Combining a mouse model of hypertension and quantitative assessment of the density of SHG signal from collagen fibers, we observed early development of interstitial fibrosis in experimental hypertension with a pattern similar to that observed in human kidney transplants. SHG signal unambiguously evidenced accumulation of collagen fibers in the tubular interstitium and in the Bowman's capsule in the kidneys from hypertensive mice and from patients with chronic renal allograft nephropathy. We also found similar localization of fibers in the interstitium and the Bowman's capsule in a mouse model of crescentic glomerulonephritis, suggesting that common pathways are recruited whatever the cause of the nephropathy.
Conversely, no SHG signal was observed in the glomerulus tuft, indicating that sclerosis of the glomerulus revealed by histological staining is induced by the accumulation of nonfibrillar collagens and other extracellular matrix compounds. Since histological and multiphoton images provide complementary information, combining the two approaches enabled the observation that glomerulosclerosis and renal fibrosis correspond to accumulation of distinct extracellular proteins. Our observation in human and mouse kidneys do not confirm the relevance of numerous reports of procollagen expression by glomerular cells in vitro, namely mesangial cells,34, 35, 36 or of high transcriptional activity of an exogenous proalpha2(I) collagen I promoter in glomeruli from hypertensive transgenic mice.37 Furthermore, our findings using SHG microscopy correlate well with immunohistochemical and immunogold labeling studies that led to a similar conclusion in human focal and segmental glomerulosclerosis, membranous glomerulonephritis, and diabetic glomerulopathy.38, 39, 40 Indeed, focal accumulation of collagen type I or III is exceptionally reported and only in severely sclerotic and obsolescent glomeruli.41, 42 Altogether, our findings suggest strong inhibition of translation of procollagen in glomerular cells in vivo. An alternative explanation could be that pathologic glomeruli may produce immunochemistry-silent and SHG-silent collagen I that is rapidly degraded or is assembled in a nonfibrillar fashion, an intriguing possibility not yet demonstrated so far, and beyond the scope of the present study.
The relationship between progression of arterial fibrosis and interstitial fibrosis is not well understood so far. High resolution images in kidneys from hypertensive mice showed striking continuity between perivascular, periglomerular, and peritubular fibers. This suggests that hypertensive renal fibrosis results from multiple cellular origins and is unlikely exclusively promoted by periarterial fibrosis or by EMT alone.
Peripheral resistance is determined at the level of the precapillary vessels. The elevated resistance in hypertensive patients is related to rarefaction (decrease in number of parallel-connected vessels) and narrowing of the lumen of resistance vessels. However, although perfusion of the kidney is a critical determinant of blood pressure regulation, very few quantitative data of remodeling of renal arterioles are available. Our normalized measurements of the arterial sizes demonstrated that the arterial adventitia do not get thicker in AngII-infused mice. This might be due to a lack of sensitivity in the measurement of minor relative changes, since arcuate arteries constitutively exhibit collagen fibers within the adventitia. However, a seven-week regime of severe hypertension that induced significant renal fibrosis and glomerulosclerosis did not affect the geometry of the adventitia of these arteries, otherwise significantly remodeled. The lack of increase of adventitia thickness contrasts with increased accumulation of interstitial fibers. This suggests that the formation of collagen fibers in the arterial adventitia and in the renal interstitium could be regulated by different pathways.
Cross sectional areas remained constant, but with increased media thickness, resulting in a decrease of the lumen radius. This observation is consistent with hypertensive inward eutrophic remodeling found in extrarenal larger arteries, such as in gluteal biopsies from hypertensive subjects and mesenteric arteries from animals that uniformly revealed reduced lumen areas and increased media-lumen ratios without an increase in medial area in resistance vessels.32, 33, 43
SHG imaging of kidney tissue also revealed high heterogeneity of collagen fibers accumulation. The maximal relative increase of SHG density associated with interstitial fibrosis measured in hypertensive mice was observed in the cortex. However, we observed scarce and thin collagen fibers, and SHG scores in this region rarely exceeded 1%, indicating that less than 1% of the tissue was invaded by collagen fibers. Scanning of large tissue volumes along with automated segmentation and scoring procedures was therefore mandatory to obtain robust data, and proved essential for accurate assessment of fibrosis.
SHG microscopy with automated scanning and image processing produced continuous scores of interstitial fibrosis that were compared with semiquantitative histopathological scoring, taking into account extracellular matrix, tubular damage, and inflammatory infiltrates, all indices being sensitive to pathologist interpretation. As expected, both scoring procedures showed a good correlation, since they measured the nephropathy progression in the same renal compartment. To determine the sensitivity threshold of cortical SHG scores, we also studied female control mice, since females are known to be less prone to fibrosis than males. The averaged cortical SHG score was: for female control mice, which is smaller than the value obtained for male control mice: . SHG scores therefore appeared to be more sensitive at early stages when Masson’s trichrome stained section revealed no apparent damages. Moreover, SHG microscopy could localize fibers and their respective extension and connections.
SHG cortical scores enabled the determination of kidney fibrosis in the same region as that used for measuring conventional glomerulosclerosis scores. Both scores showed good agreement, and we obtained a better correlation than for SHG scores calculated in the whole images. This suggests that partially common mechanisms may promote both processes, at least in a model of primary vascular injury as in AngII-infused animals.
In conclusion, this study emphasized that renal fibrosis is a markedly heterogeneous process within the kidney. We therefore developed large volume SHG imaging and relevant segmentation of the images, along with automated procedures to ensure high reproducibility and unbiased analysis. Coupling of SHG and 2PEF microscopy also proved to be a precise tool to assess the remodeling of resistive arteries in the kidney. Interestingly, AngII-induced hypertension provoked eutrophic remodeling of arcuate arteries, with no significant increase in the constitutional collagenous adventitia. We found unambiguous evidence of interstitial accumulation of fibrillar collagens in the cortex with no glomerular sclerosis in experimental severe hypertension. This surprising result suggests early onset of interstitial disease, despite causing vascular stress and distinct regulation of collagen fiber deposition in these structures. Since interstitial fibrosis occurs in a wide array of nephropathies, SHG microscopy should help pathophysiological studies and may become a diagnostic and prognostic tool for human tissues.
This work was supported by Agence Nationale de la Recherche Grant «Transcola» and by funding from the Institut National de la Santé et de la Recherche Médicale, the Centre National de la Recherche Scientifique, and Ecole Polytechnique. Strupler was supported by a grant from “Région Ile de France.” Hernest was supported by a fellowship from the French Ministry for Research and Higher Education. We thank Eric Rondeau for providing us with the human kidney biopsies and Laurent Mesnard for providing us with antimouse GBM sheep serum. We acknowledge administrative support from Martine Autran, Laure Lachapelle, and Christelle Français, and technical support from Xavier Solinas and Jean-Marc Sintès. We thank Emmanuel Beaurepaire for the outstanding design of the multiphoton setup.